It reads the place.
Live, from real sources: weather and air, sun, moon and daylight, the season and what is flowering or on the move, the wildlife people have recently seen nearby, rivers, and the ground underfoot.
pointmoon · apache-2.0
What is happening outside, at any place, right now.
Give Pointmoon a place. It reads what is going on there right now: the weather, the light, the sky, the season, the wildlife seen nearby, the water and the ground. It turns that into one clear picture and points to what is worth noticing. Every fact says where it came from and how fresh it is. When the data can’t back something, Pointmoon says so instead of guessing.
Sourced, time-stamped, or silent.
How it works
From a point on the map to what is worth noticing.
It builds a picture.
The readings become one picture of the place, arranged by what a person standing there could see, hear and feel. Anything the data cannot back is left out, not filled in.
It points to what matters now.
From that picture, Pointmoon picks out what is worth noticing at this spot at this hour: birds likely to be busy after the rain, the moon once it is actually above the horizon. When nothing stands out, it says so.
Pointmoon doesn’t write the sentence. It hands your app the facts and the moment. Your app decides what to say, and every word it says can be traced back to a real reading.
Where it runs
Built into the things we make.
Rewyld
Guided time outdoors. Rewyld reads Pointmoon so each practice meets the real weather, light and season where you are today.
Nature Class
Outdoor lessons for children. Nature Class reads Pointmoon to ground each lesson in what is around the school today: the weather, the season, and the plants and animals seen nearby.
Naturate portals, coming next
Living pages for nature places and the people who look after them, from our design studio Naturate. Each one will show what is happening at that place today, read from Pointmoon.
All three are ours. The built-with page shows exactly what each shipped product reads, with the file that proves it.
One read, many signals
One call covers the whole picture.
A single request returns every signal for the same moment, lined up on one clock. You don’t stitch six APIs together and reconcile their timestamps.
time & daylightplaceseason & what’s in bloomweather & airsky & moonwildlife seen nearbybird & pollinator activityrivers & waterthe ground underfoothow today compares to usual
Newer signals include how busy birds and pollinators are likely to be right now, with the conditions behind that estimate; river levels that say plainly when a region’s gauges are running late; and several photographs for each species seen nearby, each credited to the person who found it.
Some signals are thin in much of the world. Where they are, Pointmoon marks them silent rather than inventing them. You get what the data can actually stand behind.
The promise
So what your app says about a place is true.
Every fact is sourced.
Each fact names the feed it came from, when it was measured, and how confident that source is. If your app acts on it, it can show exactly where it came from.
Every fact has a shelf life.
Each fact says how long it stays good. Once that window is up, Pointmoon stops repeating it instead of passing an old reading off as a fresh one.
It tells you what it doesn’t know.
When the data is too thin, too old, or too uncertain, Pointmoon returns silence with a reason instead of a confident guess. It won’t always have an answer. It won’t give you a made-up one.
The proof
Grounded or silent. Measured, and live.
See it on your own coordinate.
The best proof isn’t a number on this page, it’s the live read. The /now page shows the real current conditions for your location, every value stamped with its source, its freshness, and its confidence, and every signal the data can’t back shown as silence with a reason. It’s the product, working, on your screen.
Grounded or silent, every time.
Pointmoon ships with a benchmark you can re-run, across six locations picked to stress it: a data-rich city, a coast, a high latitude, a sparse spot, a source mid-outage, and a deliberately fake coordinate. Across all six, every answer was either sourced and fresh, or an honest silence. Never a confident wrong answer.
100%grounded-or-silent
0confidently-wrong facts
6stress coordinates
Without grounding
Ask a bare model for the temperature, sky, and wind and it gives you confident numbers. No source, no timestamp. Made up, and impossible to tell apart from a real reading.
With Pointmoon
The same model answers from real facts, or says plainly what it can’t confirm. Same model, same question. One invents the world, the other reports it.
How it compares to a weather API
It sits on top of the feeds and adds what they leave out.
A weather APIPointmoon
ReturnsA numberA number, plus where it came from, when, and how sure
When data is thinA guess, or a 200 that looks fineSilence, with a reason
When a source is downHangs, errors, or quietly serves stale dataA bounded answer, honest about what’s missing, never a silent failure
ScopeOne feedWeather, air, sky, light, life, and place in one read
Built forA person reading a forecastAn app that acts on the fact and can show its work
Pointmoon isn’t competing with the feeds. It sits on top of them and adds what they leave out: where each fact came from, how fresh it is, how sure it is, and the discipline to stay quiet when it doesn’t know. There’s a backup weather source, a bounded worst-case response time, and an honest fallback when the live sources go dark. It degrades and tells you. It doesn’t hang.
Built to be called
One call, over HTTP or MCP.
Pointmoon is Apache-2.0 and made to be called by software. Reach it over plain HTTP at /api/moon?audience=facts, or add it as an MCP tool (field_truth) that any MCP client can call. A model can use it from the schema alone. The live /now read uses the same endpoint.
Four shipped products already do it, and each one is written up with the surface it powers, the exact response paths it reads, the file in its own repo that proves the integration, and a real captured payload. They are all ours. There are no outside adopters yet, and that page says so.
The exact request, the response shape, and the MCP setup are all in the docs: the quickstart to make your first call, the API reference for the contract, and the MCP server guide to wire it as a tool.
Give your app a real sense of place.
If your app says anything about the world outside, Pointmoon is how it knows. See the live read for your own location, then wire it in with one call.
Sourced, time-stamped, or silent.